< + > Three Ways AI Is Transforming Medical Coding to Improve U.S. Healthcare

The following is a guest article by Cathy Donohue, Senior Vice President of Product at CodaMetrix

For decades, niche automation and piecemeal software solutions have been deployed by health systems to make electronic health records (EHRs) more effective in delivering care, monitoring quality, and controlling costs. Much of these objectives rely on the presence of accurate code sets to describe the patients’ condition and the care delivered.

But healthcare staffing shortages combined with the growing complexity of medical coding and payer rules, and pressure on providers for more personalized care have left many hospitals swimming upstream, struggling with staffing shortages, provider burnout, and unreimbursed care.

Fortunately, in response to this turbulence, we are seeing the emergence of a new wave of next-generation artificial intelligence (AI) and deep machine learning. These solutions can deliver unprecedented accuracy and validation of claims by aggregating disparate pieces of clinical information found across the patient journey. New technology is now allowing a bedside encounter to be accurately captured and autonomously coded, transforming the providers’ workflows in ways that get claims approved faster at a lower cost.

I have spent the last 25 years focused on driving efficiency and usability by refining how IT systems interact with electronic records, all while remaining cognizant that the EHR was originally designed for registration and billing, not for physician workflows and improving patient outcomes.

The ability of AI to bolster the bottom line of health systems – by improving the speed, efficiency, and accuracy of insurance claims – is now widely understood across the healthcare sector. AI can be adapted to health systems’ existing interfaces with the help of data ingestion and integration tools, so hospitals are not reinventing the wheel to take advantage of automation to improve their bottom line.

Advanced machine learning holds enormous potential to bring order to the chaos of the modern health system. To that end, here are three ways AI is transforming medical coding to improve the financial health of practices and the quality of care accessed by patients.

Alleviate Provider Burden

Beyond the strains of patient care delivery, the status quo asks providers to spend extensive administrative time on chart documentation and medical coding. Given there are no coding classes in medical school, it’s not surprising many doctors simply default to their most used, or “favorite” codes. All too often this leads to poorly calibrated and inaccurate coding, both in respect to the acuity of a patient and the level of care provided.

Automation can now take this burden off the provider and convert that administrative time to patient care. AI has the capability to mine the medical record and historical patient timeline for a more accurate view of a patient’s diagnosis and treatment plan, saving the provider time, providing more accurate coding, and getting paid accurately for the treatment administered. There should however remain a safety net for medical coder involvement in complex cases, as well as routine human quality audits and feedback opportunities.

Improve Claim Quality

AI is able to populate the diagnostic and procedure codes reflecting the clinical specificity expressed within the EHR and apply those to the revenue cycle’s objectives for meeting the highest quality requirements of the claims process. Providers can be prescriptive in the level of quality each automated claim must meet. In the case where AI produces predictions not meeting said standard, or if claim edits are hit, cases can be routed directly to a manual coder, along with the autonomously predicted code sets, pertinent patient details, and documentation of the patient’s encounter to make the case review and coding as efficient as possible.

On both fronts, AI’s goal is to increase the speed, comprehensiveness, and accuracy of the submitted claim. The added benefit comes with the rounds of optimization the models will undergo as they learn from the human coding activity to continuously improve coverage and accuracy.

Produce Clinically Comprehensive Code Sets

While most EHRs have fulfilled their role as collectors of health systems’ digitized data, there remains a wide gap in their ability to package and display that data in a meaningful and efficient way to drive a comprehensive, longitudinal assessment of any given patient’s clinical condition.

Health systems invest millions of dollars to derive accurate assessments, in the form of diagnosis and procedural codes, for a multitude of purposes and beyond those of producing billable claims. In fact, the cost of coding across the enterprise is nearly double what is spent by the revenue cycle department.

But, if AI is able to evaluate the longitudinal record, then why stop at only producing the encounter-specific codes that meet the relatively low claim’s threshold of “medical necessity”? Instead, AI-driven autonomous coding can raise the bar and establish a clinically comprehensive set of codes, thereby supporting prior authorizations and utilization management, identifying care gaps, building care plans, and populating care registries, as well as supporting clinical research through patient recruitment for clinical trials.

It isn’t hard to take this one step further and suggest that AI could assist providers at the point of care by building an accurate and comprehensively coded problem list and encounter history. Providers would then be freed from scouring a patient’s chart to help inform the appropriate delivery of care. This can be particularly effective in emergency and bedside services, some of the most difficult service lines due to the multiplicity and complexity of diagnostic and procedure codes.

These types of transformative changes are starting to take shape in the minds of many health leaders as they begin to digest, implement, trust, and embrace the power that AI in medical coding can have across the enterprise. As understanding and confidence builds in the advantages a well-conceived AI platform can offer, deep learning will make it possible for players across complex health systems to care more and code less. That’s a future that should appeal to revenue cycle managers, providers, and patients alike.

About Cathy Donohue

Cathy Donohue is the Senior Vice President of Product at CodaMetrix, a Boston-based SaaS company that leverages AI to transform clinical data into accurate medical codes, enhancing revenue cycle management and patient care. With over a decade of experience in product strategy and operational leadership, she has successfully driven initiatives at companies like Commure and PatientKeeper, managing multimillion-dollar product portfolios and large cross-functional teams. Cathy is known for her expertise in agile development, customer relationship management, and regulatory compliance. She holds an MBA in Healthcare Administration from Boston University and a BA in Business Economics from UC Santa Barbara.